We just released Metis Agent v0.6.1, a major update that brings full local model support—just as OpenAI dropped their first open-weight models on Hugging Face (GPT OSS-117B and 21B).

With this release, you can run GPT OSS locally—no API keys, no per-token costs, no external dependencies.

What’s new in v0.6.1:

Native HuggingFace integration (including GPT OSS)

Automatic hardware detection (CUDA, MPS, CPU)

4-bit and 8-bit quantization support

36+ production-ready tools for building AI agents

One command to switch from API-based to local models

This lets developers and startups:

Build with OpenAI-quality models without usage limits

Deploy behind firewalls with full data sovereignty

Experiment and scale on local hardware.

Install Command: pip install metis-agent